Re: [phenixbb] Question about RMSD vs e/A^3

In simple terms, e-/A^3 is an absolute scale, whereas RMSD is normalized. Importantly, in an Fo-Fc map, the quantity by which you are normalizing (the RMS value of the map) is decreasing over the course of the refinement. I.e., at early stages, when the model is not optimized, there will be lots of large positive and negative peaks, many of which will disappear or at least be reduced over the course of refinement. For a perfect model (and perfect data), the RMS value of the Fo-Fc map would go to zero when the refinement is finished. For this reason, it seems to me that when comparing two different structures, sticking to an absolute scale is the safer bet, at least for the Fo-Fc maps (note that actually getting the absolute scale correct is a non-trivial problem, but that’s a discussion for another time). Also, if you’re using the same software to do the calculation in both cases, systematic problems should cancel out… Colleagues, please correct me if I’m spouting nonsense. Pat

On 30 Mar 2022, at 10:01 AM, Vatsal Purohit wrote:

Hi all,

I had a question about comparing structures based on RMSD comparisons vs e-/A^3. While the former seems to be more frequently used in structural studies with a standard value of 1.0 rmsd for 2Fo-Fc and 3.0 rmsd for Fo-Fc, there are differences in RMSD based on resolution. Hence, I wanted to ask if e-/A^3 could be a more appropriate comparison for looking at differences in electron density in a ligand-binding site between 2 structures of the same enzyme? So, here you would try and set e-/A^3 of all the maps (2Fo-Fc and Fo-Fc) to the same value.

The structures I’m referring to are between 2-2.5 angstroms in resolution.

I have also worked with the map sigma level comparison tool on PHENIX and wanted to know if that might be an alternate way to compare maps in 2 structures accordingly.

Regards, Vatsal

Post-doctoral fellow, Davis Lab, Emory University [email protected] | 346-719-9409

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe: [email protected]